The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill clones remote GitHub repositories into the /tmp directory and proceeds to run npm install and npm test. This behavior allows the execution of arbitrary code hosted in remote repositories and their dependencies. It also relies on npx to download and run the claude-flow tool at runtime.
[COMMAND_EXECUTION]: The skill frequently uses the Bash() execution environment to perform administrative tasks, such as querying the GitHub API, cloning repositories, and performing batch Git operations. These commands are constructed using variables derived from repository metadata and search results.
[INDIRECT_PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it processes untrusted data from external sources without sufficient safeguards.
Ingestion points: Fetches repository lists, search results, and file contents (package.json, CLAUDE.md) using the GitHub CLI and API.
Boundary markers: Absent. The skill does not use delimiters or instructions to prevent the agent from following commands embedded in the retrieved data.
Capability inventory: Includes arbitrary shell execution (Bash()), file writing to remote repositories (mcp__github__push_files), and creating new GitHub repositories.
Sanitization: Absent. The data retrieved from GitHub is parsed and used directly in agent coordination tasks.
[EXTERNAL_DOWNLOADS]: Automatically downloads and executes code via npx claude-flow and gh repo clone operations targeting external GitHub repositories.
[DYNAMIC_EXECUTION]: Generates and deploys new CI/CD workflows and configuration files to repositories at runtime. It also dynamically builds Pull Request descriptions based on the output of previous automated steps.

github-multi-repo